From nobody Sun May 12 01:15:34 2024 Received: from us-smtp-delivery-124.mimecast.com (us-smtp-delivery-124.mimecast.com [170.10.129.124]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id B11E28C19 for ; Wed, 11 Oct 2023 07:21:36 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=redhat.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=redhat.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (1024-bit key) header.d=redhat.com header.i=@redhat.com header.b="aOQ1MEX0" DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=redhat.com; s=mimecast20190719; t=1697008895; h=from:from:reply-to:subject:subject:date:date:message-id:message-id: to:to:cc:cc:mime-version:mime-version: content-transfer-encoding:content-transfer-encoding; bh=N41JNWYX/Pomq5SxJ5nZpshtiEKaKL8hH4wYefsbCLM=; b=aOQ1MEX0RHAh1B9tAprtAtKsjZV+vcPan5aoR8bKlJiCxXpGfXGgfLBUwr2ruPqwvj3ush ofXe4J1eBAySufTQuAJj3S2tuXK73SgDCvsgp9B786MHvO8X72YvzPXZhFn0KM906PMqmz tGm8fsYE7k4sM+9uQkg1spHZ/DcBGI0= Received: from mimecast-mx02.redhat.com (mx-ext.redhat.com [66.187.233.73]) by relay.mimecast.com with ESMTP with STARTTLS (version=TLSv1.2, cipher=TLS_ECDHE_RSA_WITH_AES_256_GCM_SHA384) id us-mta-63-4TXPEkrhMeSzJaupWnPjSQ-1; Wed, 11 Oct 2023 03:21:16 -0400 X-MC-Unique: 4TXPEkrhMeSzJaupWnPjSQ-1 Received: from smtp.corp.redhat.com (int-mx01.intmail.prod.int.rdu2.redhat.com [10.11.54.1]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mimecast-mx02.redhat.com (Postfix) with ESMTPS id 6EF2D2825E98; Wed, 11 Oct 2023 07:21:15 +0000 (UTC) Received: from gerbillo.redhat.com (unknown [10.45.225.21]) by smtp.corp.redhat.com (Postfix) with ESMTP id 960681B94; Wed, 11 Oct 2023 07:21:13 +0000 (UTC) From: Paolo Abeni To: netdev@vger.kernel.org Cc: Ayush Sawal , "David S. Miller" , Eric Dumazet , Jakub Kicinski , David Ahern , mptcp@lists.linux.dev, Boris Pismenny , Tom Deseyn Subject: [PATCH v2 net] tcp: allow again tcp_disconnect() when threads are waiting Date: Wed, 11 Oct 2023 09:20:55 +0200 Message-ID: Precedence: bulk X-Mailing-List: mptcp@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable X-Scanned-By: MIMEDefang 3.4.1 on 10.11.54.1 Content-Type: text/plain; charset="utf-8" As reported by Tom, .NET and applications build on top of it rely on connect(AF_UNSPEC) to async cancel pending I/O operations on TCP socket. The blamed commit below caused a regression, as such cancellation can now fail. As suggested by Eric, this change addresses the problem explicitly causing blocking I/O operation to terminate immediately (with an error) when a concurrent disconnect() is executed. Instead of tracking the number of threads blocked on a given socket, track the number of disconnect() issued on such socket. If such counter changes after a blocking operation releasing and re-acquiring the socket lock, error out the current operation. Fixes: 4faeee0cf8a5 ("tcp: deny tcp_disconnect() when threads are waiting") Reported-by: Tom Deseyn Closes: https://bugzilla.redhat.com/show_bug.cgi?id=3D1886305 Suggested-by: Eric Dumazet Signed-off-by: Paolo Abeni Reviewed-by: Eric Dumazet --- v1 -> v2: - move the sk_disconnects increment in __inet_stream_connect() (Eric) --- .../chelsio/inline_crypto/chtls/chtls_io.c | 36 +++++++++++++++---- include/net/sock.h | 10 +++--- net/core/stream.c | 12 ++++--- net/ipv4/af_inet.c | 10 ++++-- net/ipv4/inet_connection_sock.c | 1 - net/ipv4/tcp.c | 16 ++++----- net/ipv4/tcp_bpf.c | 4 +++ net/mptcp/protocol.c | 7 ---- net/tls/tls_main.c | 10 ++++-- net/tls/tls_sw.c | 19 ++++++---- 10 files changed, 80 insertions(+), 45 deletions(-) diff --git a/drivers/net/ethernet/chelsio/inline_crypto/chtls/chtls_io.c b/= drivers/net/ethernet/chelsio/inline_crypto/chtls/chtls_io.c index 5fc64e47568a..d567e42e1760 100644 --- a/drivers/net/ethernet/chelsio/inline_crypto/chtls/chtls_io.c +++ b/drivers/net/ethernet/chelsio/inline_crypto/chtls/chtls_io.c @@ -911,7 +911,7 @@ static int csk_wait_memory(struct chtls_dev *cdev, struct sock *sk, long *timeo_p) { DEFINE_WAIT_FUNC(wait, woken_wake_function); - int err =3D 0; + int ret, err =3D 0; long current_timeo; long vm_wait =3D 0; bool noblock; @@ -942,10 +942,13 @@ static int csk_wait_memory(struct chtls_dev *cdev, =20 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); sk->sk_write_pending++; - sk_wait_event(sk, ¤t_timeo, sk->sk_err || - (sk->sk_shutdown & SEND_SHUTDOWN) || - (csk_mem_free(cdev, sk) && !vm_wait), &wait); + ret =3D sk_wait_event(sk, ¤t_timeo, sk->sk_err || + (sk->sk_shutdown & SEND_SHUTDOWN) || + (csk_mem_free(cdev, sk) && !vm_wait), + &wait); sk->sk_write_pending--; + if (ret < 0) + goto do_error; =20 if (vm_wait) { vm_wait -=3D current_timeo; @@ -1348,6 +1351,7 @@ static int chtls_pt_recvmsg(struct sock *sk, struct m= sghdr *msg, size_t len, int copied =3D 0; int target; long timeo; + int ret; =20 buffers_freed =3D 0; =20 @@ -1423,7 +1427,11 @@ static int chtls_pt_recvmsg(struct sock *sk, struct = msghdr *msg, size_t len, if (copied >=3D target) break; chtls_cleanup_rbuf(sk, copied); - sk_wait_data(sk, &timeo, NULL); + ret =3D sk_wait_data(sk, &timeo, NULL); + if (ret < 0) { + copied =3D copied ? : ret; + goto unlock; + } continue; found_ok_skb: if (!skb->len) { @@ -1518,6 +1526,8 @@ static int chtls_pt_recvmsg(struct sock *sk, struct m= sghdr *msg, size_t len, =20 if (buffers_freed) chtls_cleanup_rbuf(sk, copied); + +unlock: release_sock(sk); return copied; } @@ -1534,6 +1544,7 @@ static int peekmsg(struct sock *sk, struct msghdr *ms= g, int copied =3D 0; size_t avail; /* amount of available data in current skb */ long timeo; + int ret; =20 lock_sock(sk); timeo =3D sock_rcvtimeo(sk, flags & MSG_DONTWAIT); @@ -1585,7 +1596,12 @@ static int peekmsg(struct sock *sk, struct msghdr *m= sg, release_sock(sk); lock_sock(sk); } else { - sk_wait_data(sk, &timeo, NULL); + ret =3D sk_wait_data(sk, &timeo, NULL); + if (ret < 0) { + /* here 'copied' is 0 due to previous checks */ + copied =3D ret; + break; + } } =20 if (unlikely(peek_seq !=3D tp->copied_seq)) { @@ -1656,6 +1672,7 @@ int chtls_recvmsg(struct sock *sk, struct msghdr *msg= , size_t len, int copied =3D 0; long timeo; int target; /* Read at least this many bytes */ + int ret; =20 buffers_freed =3D 0; =20 @@ -1747,7 +1764,11 @@ int chtls_recvmsg(struct sock *sk, struct msghdr *ms= g, size_t len, if (copied >=3D target) break; chtls_cleanup_rbuf(sk, copied); - sk_wait_data(sk, &timeo, NULL); + ret =3D sk_wait_data(sk, &timeo, NULL); + if (ret < 0) { + copied =3D copied ? : ret; + goto unlock; + } continue; =20 found_ok_skb: @@ -1816,6 +1837,7 @@ int chtls_recvmsg(struct sock *sk, struct msghdr *msg= , size_t len, if (buffers_freed) chtls_cleanup_rbuf(sk, copied); =20 +unlock: release_sock(sk); return copied; } diff --git a/include/net/sock.h b/include/net/sock.h index b770261fbdaf..92f7ea62a915 100644 --- a/include/net/sock.h +++ b/include/net/sock.h @@ -336,7 +336,7 @@ struct sk_filter; * @sk_cgrp_data: cgroup data for this cgroup * @sk_memcg: this socket's memory cgroup association * @sk_write_pending: a write to stream socket waits to start - * @sk_wait_pending: number of threads blocked on this socket + * @sk_disconnects: number of disconnect operations performed on this sock * @sk_state_change: callback to indicate change in the state of the sock * @sk_data_ready: callback to indicate there is data to be processed * @sk_write_space: callback to indicate there is bf sending space availa= ble @@ -429,7 +429,7 @@ struct sock { unsigned int sk_napi_id; #endif int sk_rcvbuf; - int sk_wait_pending; + int sk_disconnects; =20 struct sk_filter __rcu *sk_filter; union { @@ -1189,8 +1189,7 @@ static inline void sock_rps_reset_rxhash(struct sock = *sk) } =20 #define sk_wait_event(__sk, __timeo, __condition, __wait) \ - ({ int __rc; \ - __sk->sk_wait_pending++; \ + ({ int __rc, __dis =3D __sk->sk_disconnects; \ release_sock(__sk); \ __rc =3D __condition; \ if (!__rc) { \ @@ -1200,8 +1199,7 @@ static inline void sock_rps_reset_rxhash(struct sock = *sk) } \ sched_annotate_sleep(); \ lock_sock(__sk); \ - __sk->sk_wait_pending--; \ - __rc =3D __condition; \ + __rc =3D __dis =3D=3D __sk->sk_disconnects ? __condition : -EPIPE; \ __rc; \ }) =20 diff --git a/net/core/stream.c b/net/core/stream.c index f5c4e47df165..96fbcb9bbb30 100644 --- a/net/core/stream.c +++ b/net/core/stream.c @@ -117,7 +117,7 @@ EXPORT_SYMBOL(sk_stream_wait_close); */ int sk_stream_wait_memory(struct sock *sk, long *timeo_p) { - int err =3D 0; + int ret, err =3D 0; long vm_wait =3D 0; long current_timeo =3D *timeo_p; DEFINE_WAIT_FUNC(wait, woken_wake_function); @@ -142,11 +142,13 @@ int sk_stream_wait_memory(struct sock *sk, long *time= o_p) =20 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags); sk->sk_write_pending++; - sk_wait_event(sk, ¤t_timeo, READ_ONCE(sk->sk_err) || - (READ_ONCE(sk->sk_shutdown) & SEND_SHUTDOWN) || - (sk_stream_memory_free(sk) && - !vm_wait), &wait); + ret =3D sk_wait_event(sk, ¤t_timeo, READ_ONCE(sk->sk_err) || + (READ_ONCE(sk->sk_shutdown) & SEND_SHUTDOWN) || + (sk_stream_memory_free(sk) && !vm_wait), + &wait); sk->sk_write_pending--; + if (ret < 0) + goto do_error; =20 if (vm_wait) { vm_wait -=3D current_timeo; diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c index 3d2e30e20473..2713c9b06c4c 100644 --- a/net/ipv4/af_inet.c +++ b/net/ipv4/af_inet.c @@ -597,7 +597,6 @@ static long inet_wait_for_connect(struct sock *sk, long= timeo, int writebias) =20 add_wait_queue(sk_sleep(sk), &wait); sk->sk_write_pending +=3D writebias; - sk->sk_wait_pending++; =20 /* Basic assumption: if someone sets sk->sk_err, he _must_ * change state of the socket from TCP_SYN_*. @@ -613,7 +612,6 @@ static long inet_wait_for_connect(struct sock *sk, long= timeo, int writebias) } remove_wait_queue(sk_sleep(sk), &wait); sk->sk_write_pending -=3D writebias; - sk->sk_wait_pending--; return timeo; } =20 @@ -642,6 +640,7 @@ int __inet_stream_connect(struct socket *sock, struct s= ockaddr *uaddr, return -EINVAL; =20 if (uaddr->sa_family =3D=3D AF_UNSPEC) { + sk->sk_disconnects++; err =3D sk->sk_prot->disconnect(sk, flags); sock->state =3D err ? SS_DISCONNECTING : SS_UNCONNECTED; goto out; @@ -696,6 +695,7 @@ int __inet_stream_connect(struct socket *sock, struct s= ockaddr *uaddr, int writebias =3D (sk->sk_protocol =3D=3D IPPROTO_TCP) && tcp_sk(sk)->fastopen_req && tcp_sk(sk)->fastopen_req->data ? 1 : 0; + int dis =3D sk->sk_disconnects; =20 /* Error code is set above */ if (!timeo || !inet_wait_for_connect(sk, timeo, writebias)) @@ -704,6 +704,11 @@ int __inet_stream_connect(struct socket *sock, struct = sockaddr *uaddr, err =3D sock_intr_errno(timeo); if (signal_pending(current)) goto out; + + if (dis !=3D sk->sk_disconnects) { + err =3D -EPIPE; + goto out; + } } =20 /* Connection was closed by RST, timeout, ICMP error @@ -725,6 +730,7 @@ int __inet_stream_connect(struct socket *sock, struct s= ockaddr *uaddr, sock_error: err =3D sock_error(sk) ? : -ECONNABORTED; sock->state =3D SS_UNCONNECTED; + sk->sk_disconnects++; if (sk->sk_prot->disconnect(sk, flags)) sock->state =3D SS_DISCONNECTING; goto out; diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_soc= k.c index aeebe8816689..394a498c2823 100644 --- a/net/ipv4/inet_connection_sock.c +++ b/net/ipv4/inet_connection_sock.c @@ -1145,7 +1145,6 @@ struct sock *inet_csk_clone_lock(const struct sock *s= k, if (newsk) { struct inet_connection_sock *newicsk =3D inet_csk(newsk); =20 - newsk->sk_wait_pending =3D 0; inet_sk_set_state(newsk, TCP_SYN_RECV); newicsk->icsk_bind_hash =3D NULL; newicsk->icsk_bind2_hash =3D NULL; diff --git a/net/ipv4/tcp.c b/net/ipv4/tcp.c index 3f66cdeef7de..d3456cf840de 100644 --- a/net/ipv4/tcp.c +++ b/net/ipv4/tcp.c @@ -831,7 +831,9 @@ ssize_t tcp_splice_read(struct socket *sock, loff_t *pp= os, */ if (!skb_queue_empty(&sk->sk_receive_queue)) break; - sk_wait_data(sk, &timeo, NULL); + ret =3D sk_wait_data(sk, &timeo, NULL); + if (ret < 0) + break; if (signal_pending(current)) { ret =3D sock_intr_errno(timeo); break; @@ -2442,7 +2444,11 @@ static int tcp_recvmsg_locked(struct sock *sk, struc= t msghdr *msg, size_t len, __sk_flush_backlog(sk); } else { tcp_cleanup_rbuf(sk, copied); - sk_wait_data(sk, &timeo, last); + err =3D sk_wait_data(sk, &timeo, last); + if (err < 0) { + err =3D copied ? : err; + goto out; + } } =20 if ((flags & MSG_PEEK) && @@ -2966,12 +2972,6 @@ int tcp_disconnect(struct sock *sk, int flags) int old_state =3D sk->sk_state; u32 seq; =20 - /* Deny disconnect if other threads are blocked in sk_wait_event() - * or inet_wait_for_connect(). - */ - if (sk->sk_wait_pending) - return -EBUSY; - if (old_state !=3D TCP_CLOSE) tcp_set_state(sk, TCP_CLOSE); =20 diff --git a/net/ipv4/tcp_bpf.c b/net/ipv4/tcp_bpf.c index 327268203001..ba2e92188124 100644 --- a/net/ipv4/tcp_bpf.c +++ b/net/ipv4/tcp_bpf.c @@ -307,6 +307,8 @@ static int tcp_bpf_recvmsg_parser(struct sock *sk, } =20 data =3D tcp_msg_wait_data(sk, psock, timeo); + if (data < 0) + return data; if (data && !sk_psock_queue_empty(psock)) goto msg_bytes_ready; copied =3D -EAGAIN; @@ -351,6 +353,8 @@ static int tcp_bpf_recvmsg(struct sock *sk, struct msgh= dr *msg, size_t len, =20 timeo =3D sock_rcvtimeo(sk, flags & MSG_DONTWAIT); data =3D tcp_msg_wait_data(sk, psock, timeo); + if (data < 0) + return data; if (data) { if (!sk_psock_queue_empty(psock)) goto msg_bytes_ready; diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c index c3b83cb390d9..d1902373c974 100644 --- a/net/mptcp/protocol.c +++ b/net/mptcp/protocol.c @@ -3098,12 +3098,6 @@ static int mptcp_disconnect(struct sock *sk, int fla= gs) { struct mptcp_sock *msk =3D mptcp_sk(sk); =20 - /* Deny disconnect if other threads are blocked in sk_wait_event() - * or inet_wait_for_connect(). - */ - if (sk->sk_wait_pending) - return -EBUSY; - /* We are on the fastopen error path. We can't call straight into the * subflows cleanup code due to lock nesting (we are already under * msk->firstsocket lock). @@ -3173,7 +3167,6 @@ struct sock *mptcp_sk_clone_init(const struct sock *s= k, inet_sk(nsk)->pinet6 =3D mptcp_inet6_sk(nsk); #endif =20 - nsk->sk_wait_pending =3D 0; __mptcp_init_sock(nsk); =20 msk =3D mptcp_sk(nsk); diff --git a/net/tls/tls_main.c b/net/tls/tls_main.c index 02f583ff9239..002483e60c19 100644 --- a/net/tls/tls_main.c +++ b/net/tls/tls_main.c @@ -139,8 +139,8 @@ void update_sk_prot(struct sock *sk, struct tls_context= *ctx) =20 int wait_on_pending_writer(struct sock *sk, long *timeo) { - int rc =3D 0; DEFINE_WAIT_FUNC(wait, woken_wake_function); + int ret, rc =3D 0; =20 add_wait_queue(sk_sleep(sk), &wait); while (1) { @@ -154,9 +154,13 @@ int wait_on_pending_writer(struct sock *sk, long *time= o) break; } =20 - if (sk_wait_event(sk, timeo, - !READ_ONCE(sk->sk_write_pending), &wait)) + ret =3D sk_wait_event(sk, timeo, + !READ_ONCE(sk->sk_write_pending), &wait); + if (ret) { + if (ret < 0) + rc =3D ret; break; + } } remove_wait_queue(sk_sleep(sk), &wait); return rc; diff --git a/net/tls/tls_sw.c b/net/tls/tls_sw.c index d1fc295b83b5..e9d1e83a859d 100644 --- a/net/tls/tls_sw.c +++ b/net/tls/tls_sw.c @@ -1291,6 +1291,7 @@ tls_rx_rec_wait(struct sock *sk, struct sk_psock *pso= ck, bool nonblock, struct tls_context *tls_ctx =3D tls_get_ctx(sk); struct tls_sw_context_rx *ctx =3D tls_sw_ctx_rx(tls_ctx); DEFINE_WAIT_FUNC(wait, woken_wake_function); + int ret =3D 0; long timeo; =20 timeo =3D sock_rcvtimeo(sk, nonblock); @@ -1302,6 +1303,9 @@ tls_rx_rec_wait(struct sock *sk, struct sk_psock *pso= ck, bool nonblock, if (sk->sk_err) return sock_error(sk); =20 + if (ret < 0) + return ret; + if (!skb_queue_empty(&sk->sk_receive_queue)) { tls_strp_check_rcv(&ctx->strp); if (tls_strp_msg_ready(ctx)) @@ -1320,10 +1324,10 @@ tls_rx_rec_wait(struct sock *sk, struct sk_psock *p= sock, bool nonblock, released =3D true; add_wait_queue(sk_sleep(sk), &wait); sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk); - sk_wait_event(sk, &timeo, - tls_strp_msg_ready(ctx) || - !sk_psock_queue_empty(psock), - &wait); + ret =3D sk_wait_event(sk, &timeo, + tls_strp_msg_ready(ctx) || + !sk_psock_queue_empty(psock), + &wait); sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk); remove_wait_queue(sk_sleep(sk), &wait); =20 @@ -1852,6 +1856,7 @@ static int tls_rx_reader_acquire(struct sock *sk, str= uct tls_sw_context_rx *ctx, bool nonblock) { long timeo; + int ret; =20 timeo =3D sock_rcvtimeo(sk, nonblock); =20 @@ -1861,14 +1866,16 @@ static int tls_rx_reader_acquire(struct sock *sk, s= truct tls_sw_context_rx *ctx, ctx->reader_contended =3D 1; =20 add_wait_queue(&ctx->wq, &wait); - sk_wait_event(sk, &timeo, - !READ_ONCE(ctx->reader_present), &wait); + ret =3D sk_wait_event(sk, &timeo, + !READ_ONCE(ctx->reader_present), &wait); remove_wait_queue(&ctx->wq, &wait); =20 if (timeo <=3D 0) return -EAGAIN; if (signal_pending(current)) return sock_intr_errno(timeo); + if (ret < 0) + return ret; } =20 WRITE_ONCE(ctx->reader_present, 1); --=20 2.41.0